Research map / Detection and segmentation / Tracking
Where and What: Long-Term Object Tracking in Egocentric Videos
Jacob Chalk, Saptarshi Sinha, Dima Damen, Yannis Kalantidis, Diane Larlus
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add appearance information (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 34.0 -> 72.2 | Improved |
| add location information (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 62.6 -> 72.2 | Improved |
| add New Track token (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 53.0 -> 72.2 | Improved |
| learn the NT token instead of fixed random (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 72.0 -> 72.2 | Improved |
| use DenStream clusters instead of most recent appearance (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 52.9 -> 72.2 | Improved |
| use DenStream instead of infinite memory for appearance (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 70.6 -> 72.2 | Improved |
| add a 1 layer transformer assignment encoder (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 66.6 -> 72.2 | Improved |
| increase encoder layers from 1 to 2 (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 72.2 -> 70.8 | Hurt |
| add depth alignment for 3D locations (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 70.5 -> 72.2 | Improved |
| train with DAgger correction instead of teacher forcing (ablation) | Whareformer · HD-EPIC | mPCL (%): 41.6 -> 43.6 | Improved |
| train with DAgger correction instead of teacher forcing (ablation) | Whareformer · EPIC-KITCHENS | mPCL (%): 72.2 -> 71.0 | Hurt |
| replace LMK with Whareformer | LMK · IT3DEgo | mPCL (%): 31.0 -> 58.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ByteTrack — ByteTrack: Multi-Object Tracking by Associating Every Detection Box (ECCV 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- EgoTracks: A Long-term Egocentric Visual Object Tracking Dataset (NeurIPS 2023 · similarity 0.91)
- Instance Tracking in 3D Scenes from Egocentric Videos (CVPR 2024 · similarity 0.90)
- Object-Shot Enhanced Grounding Network for Egocentric Video (CVPR 2025 · similarity 0.89)
- Towards Stable Self-Supervised Object Representations in Unconstrained Egocentric Video (CVPR 2026 · similarity 0.89)
- EgoThinker: Unveiling Egocentric Reasoning with Spatio-Temporal CoT (NeurIPS 2025 · similarity 0.87)
- Test-Time Adaptation for Combating Missing Modalities in Egocentric Videos (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
